Deep Learning Algorithms with LIME and Similarity Distance Analysis on COVID-19 Chest X-ray Dataset.

In the last few years, many types of research have been conducted on the most harmful pandemic, COVID-19. Machine learning approaches have been applied to investigate chest X-rays of COVID-19 patients in many respects. This study focuses on the deep learning algorithm from the standpoint of feature space and similarity analysis. Firstly, we utilized Local Interpretable Model-agnostic Explanations (LIME) to justify the necessity of the region of interest (ROI) process and further prepared ROI via U-Net segmentation that masked out non-lung areas of images to prevent the classifier from being distracted by irrelevant features. The experimental results were promising, with detection performance reaching an overall accuracy of 95.5%, a sensitivity of 98.4%, a precision of 94.7%, and an F1 score of 96.5% on the COVID-19 category. Secondly, we applied similarity analysis to identify outliers and further provided an objective confidence reference specific to the similarity distance to centers or boundaries of clusters while inferring. Finally, the experimental results suggested putting more effort into enhancing the low-accuracy subspace locally, which is identified by the similarity distance to the centers. The experimental results were promising, and based on those perspectives, our approach could be more flexible to deploy dedicated classifiers specific to different subspaces instead of one rigid end-to-end black box model for all feature space.

A spatial analysis of TB cases and abnormal X-rays detected through active case-finding in Karachi, Pakistan.

Tuberculosis (TB) is the leading cause of avoidable deaths from an infectious disease globally and a large of number of people who develop TB each year remain undiagnosed. Active case-finding has been recommended by the World Health Organization to bridge the case-detection gap for TB in high burden countries. However, concerns remain regarding their yield and cost-effectiveness. Data from mobile chest X-ray (CXR) supported active case-finding community camps conducted in Karachi, Pakistan from July 2018 to March 2020 was retrospectively analyzed. Frequency analysis was carried out at the camp-level and outcomes of interest for the spatial analyses were mycobacterium TB positivity (MTB+) and X-ray abnormality rates. The Global Moran's I statistic was used to test for spatial autocorrelation for MTB+ and abnormal X-rays within Union Councils (UCs) in Karachi. A total of 1161 (78.1%) camps yielded no MTB+ cases, 246 (16.5%) camps yielded 1 MTB+, 52 (3.5%) camps yielded 2 MTB+ and 27 (1.8%) yielded 3 or more MTB+. A total of 79 (5.3%) camps accounted for 193 (44.0%) of MTB+ cases detected. Statistically significant clustering for MTB positivity (Global Moran's I: 0.09) and abnormal chest X-rays (Global Moran's I: 0.36) rates was identified within UCs in Karachi. Clustering of UCs with high MTB positivity were identified in Karachi West district. Statistically significant spatial variation was identified in yield of bacteriologically positive TB cases and in abnormal CXR through active case-finding in Karachi. Cost-effectiveness of active case-finding programs can be improved by identifying and focusing interventions in hotspots and avoiding locations with no known TB cases reported through routine surveillance.

Chest x-ray findings of paediatric patients seen at a Teaching Hospital in North-western Nigeria

Introduction: Plain radiography is the first line of radiological examination used for the evaluation of paediatric chest, and also a rapid imaging technique that allows lung abnormality to be identified. X-ray is used to diagnose conditions of the thoracic cavity, including airways, ribs, lungs, heart, and diaphragm. Chest x-ray has a high sensitivity for pulmonary tuberculosis and thus is a valuable tool to identify a differential diagnosis for a patient. Methodology: Two hundred and eleven paediatrics chest x-ray reports were studied between February 2017 to September 2018. Data were collected retrospectively from the hospital archives using a data capture sheet. Results: Two hundred and eleven radiographs were assessed and the age of patients whose radiographs participated in the study ranged from 0-15 years. Also, a greater number of male patients 133(63.03%) participated in the study than female patients 78(36.97%). Among all the radiographs used in this study, the most common age group for this study ranged from 0-3 years. Results from the study also revealed that bronchopneumonia was the most common finding totaling 105(49.76%) followed by radiographs which are normal findings 77(36.49), pulmonary tuberculosis, 19(9.00%), congestive heart failure 4(1.90%), pleural effusion 3(1.42%), enlarged adenoid 2(0.95%) and dextrocardia 1(0.47%). Conclusion: This study report bronchopneumonia was the most common paediatrics radiographic finding in a chest x-ray. Plain radiography is an effective tool to examine various respiratory and cardiac pathologies and is the first line of investigation for chest pathologies.

VinDr-CXR: An open dataset of chest X-rays with radiologist's annotations.

Most of the existing chest X-ray datasets include labels from a list of findings without specifying their locations on the radiographs. This limits the development of machine learning algorithms for the detection and localization of chest abnormalities. In this work, we describe a dataset of more than 100,000 chest X-ray scans that were retrospectively collected from two major hospitals in Vietnam. Out of this raw data, we release 18,000 images that were manually annotated by a total of 17 experienced radiologists with 22 local labels of rectangles surrounding abnormalities and 6 global labels of suspected diseases. The released dataset is divided into a training set of 15,000 and a test set of 3,000. Each scan in the training set was independently labeled by 3 radiologists, while each scan in the test set was labeled by the consensus of 5 radiologists. We designed and built a labeling platform for DICOM images to facilitate these annotation procedures. All images are made publicly available in DICOM format along with the labels of both the training set and the test set.

Risk Factors and Clinical Impact of Fibrotic-Like Changes and the Organizing Pneumonia Pattern in Patients With COVID-19- and Non-COVID-19-Induced Acute Respiratory Distress Syndrome / Factores de riesgo e impacto clínico de los cambios fibróticos y el patrón de neumonía organizada en los pacientes con síndrome de distrés respiratorio agudo producido o no por COVID-19

No disponible

Descriptive study of chest x-ray examination in mandatory annual health examinations at the workplace in Japan.

The utility of chest x-ray examination (CXR) in mandatory annual health examinations for occupational health is debatable in Japan. This study aimed to provide basic data to consider future policies for mandatory annual health examinations in the workplace. A nationwide descriptive survey was performed to determine the rate of detection of tuberculosis, lung cancer, and other diseases through CXR in organizations associated with National Federation of Industrial Health Association. The rate of finding on CXR conducted during annual health examinations in FY2016 was evaluated. Data regarding diagnosis based on follow-up examination findings were obtained and compared with the national statistics. In addition, CXR findings were compared with the results of low-dose lung computed tomography performed at the Hitachi Health Care Center. From 121 surveyed institutions, 88 institutions with 8,669,403 workers were included. For all ages, 1.0% of examinees required follow-up examination. Among 4,764,985 workers with diagnosis data, the tuberculosis detection rate was 1.8-5.3 per 100,000 persons. For Lung cancer, 3,688,396 workers were surveyed, and 334 positive cases were detected. The lung cancer detection rate using CXR was 9.1-24.4 per 100,000 persons. From 164 cases with information regarding the clinical stage, 72 (43.9%) had Stage I lung cancer. From 40,045 workers who underwent low-dose computed tomography multiple times, 31 lung cancer cases, all with Stage I disease, were detected (detection rate: 77.4 per 100,000 persons). Our findings suggest that CXR plays a little role in the detection of active tuberculosis. With regard to LC screening, the detection rate of LC by CXR was lower, approximately 50%, than the expected rate (41.0 per 100,000 persons) of LC morbidity based on the age-sex distribution of this study population. However, the role of CXR for LC screening cannot be mentioned based on this result, because assessment of mortality reduction is essential to evaluate the role.

The impact of chest CT body composition parameters on clinical outcomes in COVID-19 patients.

We assessed the impact of chest CT body composition parameters on outcomes and disease severity at hospital presentation of COVID-19 patients, focusing also on the possible mediation of body composition in the relationship between age and death in these patients. Chest CT scans performed at hospital presentation by consecutive COVID-19 patients (02/27/2020-03/13/2020) were retrospectively reviewed to obtain pectoralis muscle density and total, visceral, and intermuscular adipose tissue areas (TAT, VAT, IMAT) at the level of T7-T8 vertebrae. Primary outcomes were: hospitalization, mechanical ventilation (MV) and/or death, death alone. Secondary outcomes were: C-reactive protein (CRP), oxygen saturation (SO2), CT disease extension at hospital presentation. The mediation of body composition in the effect of age on death was explored. Of the 318 patients included in the study (median age 65.7 years, females 37.7%), 205 (64.5%) were hospitalized, 68 (21.4%) needed MV, and 58 (18.2%) died. Increased muscle density was a protective factor while increased TAT, VAT, and IMAT were risk factors for hospitalization and MV/death. All these parameters except TAT had borderline effects on death alone. All parameters were associated with SO2 and extension of lung parenchymal involvement at CT; VAT was associated with CRP. Approximately 3% of the effect of age on death was mediated by decreased muscle density. In conclusion, low muscle quality and ectopic fat accumulation were associated with COVID-19 outcomes, VAT was associated with baseline inflammation. Low muscle quality partly mediated the effect of age on mortality.

Active case finding of pulmonary TB in a European refugee camp: lessons learnt from Oinofyta hosting site in Greece.

OBJECTIVES: To report on an active case finding (ACF) intervention that took place in the migrant camp of Oinofyta, Greece, upon suspicion of active TB transmission. METHODS: Upon diagnosis of 3 TB cases among camp residents, an ACF intervention among contacts was implemented. All camp residents were offered two-step screening, that is tuberculin skin testing (TST) followed by chest X-ray in case of positive TST (defined as ≥5 mm). RESULTS: 336 of 379 (89%) camp residents underwent TST testing, of whom 110 (33%) exhibited a positive skin reaction. The rate of positive TST results was particularly high in the elderly and significantly higher in adults than in children. Differences by sex or nationality were not observed. Of the 110 cases with positive TST, only 75 underwent chest X-ray, resulting in the detection of one pulmonary TB case in an adult woman. CONCLUSIONS: In the given intervention context, two-step ACF proved to be operationally cumbersome, with many residents lost to follow-up and a high Number Needed to Screen. Simpler ACF designs should be pilot-tested in similar settings in the future, and blanket screening of all camp residents should be reconsidered. Conclusions drawn by these exercises should pave the way for adopting a comprehensive, context-specific and evidence-based national strategy on TB in migrants.

Adapting for the COVID-19 pandemic in Ecuador, a characterization of hospital strategies and patients.

The World Health Organization (WHO) declared coronavirus disease-2019 (COVID-19) a global pandemic on 11 March 2020. In Ecuador, the first case of COVID-19 was recorded on 29 February 2020. Despite efforts to control its spread, SARS-CoV-2 overran the Ecuadorian public health system, which became one of the most affected in Latin America on 24 April 2020. The Hospital General del Sur de Quito (HGSQ) had to transition from a general to a specific COVID-19 health center in a short period of time to fulfill the health demand from patients with respiratory afflictions. Here, we summarized the implementations applied in the HGSQ to become a COVID-19 exclusive hospital, including the rearrangement of hospital rooms and a triage strategy based on a severity score calculated through an artificial intelligence (AI)-assisted chest computed tomography (CT). Moreover, we present clinical, epidemiological, and laboratory data from 75 laboratory tested COVID-19 patients, which represent the first outbreak of Quito city. The majority of patients were male with a median age of 50 years. We found differences in laboratory parameters between intensive care unit (ICU) and non-ICU cases considering C-reactive protein, lactate dehydrogenase, and lymphocytes. Sensitivity and specificity of the AI-assisted chest CT were 21.4% and 66.7%, respectively, when considering a score >70%; regardless, this system became a cornerstone of hospital triage due to the lack of RT-PCR testing and timely results. If health workers act as vectors of SARS-CoV-2 at their domiciles, they can seed outbreaks that might put 1,879,047 people at risk of infection within 15 km around the hospital. Despite our limited sample size, the information presented can be used as a local example that might aid future responses in low and middle-income countries facing respiratory transmitted epidemics.

Chest X-ray interpretation does not complement Xpert MTB/RIF in diagnosis of smear-negative pulmonary tuberculosis among TB-HIV co-infected adults in a resource-limited setting.

BACKGROUND: Chest X-ray (CXR) interpretation remains a central component of the current World Health Organization recommendations as an adjuvant test in diagnosis of smear-negative tuberculosis (TB). With its low specificity, high maintenance and operational costs, utility of CXR in diagnosis of smear-negative TB in high HIV/TB burden settings in the Xpert MTB/RIF era remains unpredictable. We evaluated accuracy and additive value of CXR to Xpert MTB/RIF in the diagnosis of TB among HIV-positive smear-negative presumptive TB patients. METHODS: HIV co-infected presumptive TB patients were recruited from the Infectious Diseases Institute outpatient clinic and in-patient medical wards of Mulago Hospital, Uganda. CXR films were reviewed by two independent radiologists using a standardized evaluation form. CXR interpretation with regard to TB was either positive (consistent with TB) or negative (normal or unlikely TB). Sensitivity, specificity and predictive values of CXR and CXR combined with Xpert MTB/RIF for diagnosis of smear-negative TB in HIV-positive patients were calculated using sputum and/or blood mycobacterial culture as reference standard. RESULTS: Three hundred sixty-six HIV co-infected smear-negative participants (female, 63.4%; hospitalized, 68.3%) had technically interpretable CXR. Median (IQR) age was 32 (28-39) years and CD4 count 112 (23-308) cells/mm3. Overall, 22% (81/366) were positive for Mycobacterium tuberculosis (Mtb) on culture; 187/366 (51.1%) had CXR interpreted as consistent with TB, of which 55 (29.4%) had culture-confirmed TB. Sensitivity and specificity of CXR interpretation in diagnosis of culture-positive TB were 67.9% (95%CI 56.6-77.8) and 53.7% (95%CI 47.7-59.6) respectively, while Xpert MTB/RIF sensitivity and specificity were 65.4% (95%CI 54.0-75.7) and 95.8% (95%CI 92.8-97.8) respectively. Addition of CXR to Xpert MTB/RIF had overall sensitivity and specificity of 87.7% (95%CI 78.5-93.9) and 51.6% (95%CI 45.6-57.5) respectively; 86.2% (95%CI 75.3-93.5) and 48.1% (95%CI 40.7-55.6) among inpatients and 93.8% (95%CI 69.8-99.8) and 58.0% (95%CI 47.7-67.8) among outpatients respectively. CONCLUSION: In this high prevalence TB/HIV setting, CXR interpretation added sensitivity to Xpert MTB/RIF test at the expense of specificity in the diagnosis of culture-positive TB in HIV-positive individuals presenting with TB symptoms and negative smear. CXR interpretation may not add diagnostic value in settings where Xpert MTB/RIF is available as a TB diagnostic tool.

Prevalence of pneumonia by chest x-ray, associated demographic characteristics and health risk factors among COVID-19 patients in Ghana

Objective: The study was conducted to determine the prevalence of radiologically diagnosed pneumonia among COVID-19 patients and associated factors. Design, setting, and participants: A retrospective manual data extraction of 275 medical records of COVID-19 patients was conducted at two COVID-19 national treatment centres in Accra from March to May 2020. All patients had a chest x-ray done. Main outcome and analysis: The main outcome was the presence of pneumonia. Descriptive statistics and Chi-square test of independence were employed to determine the associations between independent variables and the presence of pneumonia. All analysis was performed using Stata 16, and a p-value ≤ 0.05 was deemed significant Results: The prevalence of pneumonia was 44%(95%CI) =38.2-50.0). Chi-square independent test indicated that pneumonia in the COVID-19 patients was associated with educational level, history of domestic and international travel, mass gathering in the past 14 days before diagnosis, and discharge plan (p-value< 0.05). Patients classified as secondary cases (61.5%) and those discharged as fully recovered from the health facility (61.2%) had a higher prevalence of pneumonia. In addition, COVID-19 patients with hypertension (32.1%) and asthma (5.2%) had a significantly higher prevalence of pneumonia. Conclusion: Overall, the prevalence of pneumonia was 44% and was associated with the demographic and personal characteristics of the patients. Early detection through contact tracing and community surveillance should be intensified to pick up more asymptomatic cases. The role of the chest x-ray for triaging patients and for clinical management of symptomatic patients remains key

Apport de la TDM thoracique pour le dépistage de la COVID-19 et impact médico-économique de la pandémie à COVID-19 au service de radiologie de Fann (Dakar-Sénégal) / Contribution of thoracic CT for COVID-19 screening and medico-economic impact of the COVID-19 pandemic in the radiology department of Fann (Dakar-Senegal)

Objectifs :Évaluer l'apport de la TDM thoracique dans le diagnostic des patients suspects de COVID-19en comparaison avec la technique de référence (RT-PCR) et déterminer l'impact médico-économique de la COVID-19 au service de radiologie du CHU de Fann. Matériels et méthodes: Il s'agissait d'une étude rétrospective, descriptive sur une période de 4 mois allant du 1er avril au 31 Juillet 2020, au service de radiologie du CHU de Fann.Ont été inclus les patients reçus pour suspicion clinique de COVID-19, qui avaient eu une TDM thoracique et un prélèvement naso-pharyngé par écouvillonnage pour RT-PCR, soit au total 314 patients. Nous avons étudié les données épidémiologiques, cliniques, les images évocatrices de COVID-19 (opacités en verre dépoli, condensation, topographie lésionnelle), existence ou non d'une embolie, les anomalies en faveur de surinfection, les lésions associées, l'impact sur la fréquentation des différentes modalités etl'impact sur les recettes. Résultats:L'âge médian était de 62 ans et le sex-ratio 1,61. La fièvre a été présente chez 7 patients (2,23%); la toux chez 17 patients (5,41%); la dyspnée chez 30 patients (9,55%) et un syndrome de détresse respiratoire chez 63 patients (20,07%). La clinique n'a pas été précisée chez 163 patients (51,91%). La TDM thoracique était normale chez 20 patients (6,37%), évocatrice de COVID-19 chez 274 patients (87,26%) et non évocatrice de COVID-19 chez 20 patients (6,37%). La RT-PCR était positive chez 125 patients soit 39,80%. La sensibilité et la spécificité de la TDM étaient respectivement de 91,2% et 15,34%. La valeur prédictive positive, la valeur prédictive négative et le taux de précision étaient respectivement de 42%, 72,5% et 45,5%. La baisse du taux de fréquentation était de 59% en radiographie standard, 55% en échographie, 24% au scanner et 87% en mammographie. Dans notre étude on a noté une baisse de 40% des recettes au second trimestre de 2020 comparativement au premier trimestre. Conclusion: La TDM thoracique a une bonne sensibilité pour le diagnostic de la COVID-19. De ce fait, elle peut être considérée comme un outil principal pour la détection des lésions pulmonaires évocatrices de pneumonie COVID-19. Les impacts médico-économiques de la COVID-19 ont été considérables.

Slow radiological improvement and persistent low-grade inflammation after chemotherapy in tuberculosis patients with type 2 diabetes.

BACKGROUND: Diabetes mellitus type 2 (DM) may impede immune responses in tuberculosis (TB) and thus contribute to enhanced disease severity. In this study, we aimed to evaluate DM-mediated alterations in clinical, radiological and immunological outcomes in TB disease. METHODS: Newly diagnosed pulmonary TB patients with or without DM (TB n = 40; TB-DM n = 40) were recruited in Dhaka, Bangladesh. Clinical symptoms, sputum smear and culture conversion as well as chest radiography were assessed. Peripheral blood and sputum samples were collected at the time of diagnosis (baseline) and after 1, 2 and 6 months of standard anti-TB treatment. Blood samples were also obtained from healthy controls (n = 20). mRNA expression of inflammatory markers in blood and sputum samples were quantified using real-time PCR. RESULTS: The majority of TB-DM patients had poor glycemic control (HbA1c > 8%) and displayed elevated pulmonary pathology (P = 0.039) particularly in the middle (P < 0.004) and lower lung zones (P < 0.02) throughout the treatment period. However, reduction of clinical symptoms and time to sputum smear and culture conversion did not differ between the groups. Transcripts levels of the pro-inflammatory cytokines IL-1ß (P = 0.003 at month-1 and P = 0.045 at month-2) and TNF-α (P = 0.005 at month-1) and the anti-inflammatory cytokine IL-10 (P = 0.005 at month-2) were higher in peripheral blood after anti-TB treatment in TB-DM compared to TB patients. Conversely in sputum, TB-DM patients had reduced CD4 (P < 0.009 at month-1) and IL-10 (P = 0.005 at month-1 and P = 0.006 at month-2) transcripts, whereas CD8 was elevated (P = 0.016 at month-2). At 1- and 2-month post-treatment, sputum IL-10 transcripts were inversely correlated with fasting blood glucose and HbA1c levels in all patients. CONCLUSION: Insufficient up-regulation of IL-10 in the lung may fuel persistent local inflammation thereby promoting lung pathology in TB-DM patients with poorly controlled DM.

Diagnostic performance of chest computed tomography during the epidemic wave of COVID-19 varied as a function of time since the beginning of the confinement in France.

OBJECTIVE: To evaluate the diagnostic performance of the initial chest CT to diagnose COVID-19 related pneumonia in a French population of patients with respiratory symptoms according to the time from the onset of country-wide confinement to better understand what could be the role of the chest CT in the different phases of the epidemic. MATERIAL AND METHOD: Initial chest CT of 1064 patients with respiratory symptoms suspect of COVID-19 referred between March 18th, and May 12th 2020, were read according to a standardized procedure. The results of chest CTs were compared to the results of the RT-PCR. RESULTS: 546 (51%) patients were found to be positive for SARS-CoV2 at RT-PCR. The highest rate of positive RT-PCR was during the second week of confinement reaching 71.9%. After six weeks of confinement, the positive RT-PCR rate dropped significantly to 10.5% (p<0.001) and even 2.2% during the two last weeks. Overall, CT revealed patterns suggestive of COVID-19 in 603 patients (57%), whereas an alternative diagnosis was found in 246 patients (23%). CT was considered normal in 215 patients (20%) and inconclusive in 1 patient. The overall sensitivity of CT was 88%, specificity 76%, PPV 79%, and NPV 85%. At week-2, the same figures were 89%, 69%, 88% and 71% respectively and 60%, 84%, 30% and 95% respectively at week-6. At the end of confinement when the rate of positive PCR became extremely low the sensitivity, specificity, PPV and NPV of CT were 50%, 82%, 6% and 99% respectively. CONCLUSION: At the peak of the epidemic, chest CT had sufficiently high sensitivity and PPV to serve as a first-line positive diagnostic tool but at the end of the epidemic wave CT is more useful to exclude COVID-19 pneumonia.

Implementation of convolutional neural network approach for COVID-19 disease detection.

In this paper, two novel, powerful, and robust convolutional neural network (CNN) architectures are designed and proposed for two different classification tasks using publicly available data sets. The first architecture is able to decide whether a given chest X-ray image of a patient contains COVID-19 or not with 98.92% average accuracy. The second CNN architecture is able to divide a given chest X-ray image of a patient into three classes (COVID-19 versus normal versus pneumonia) with 98.27% average accuracy. The hyperparameters of both CNN models are automatically determined using Grid Search. Experimental results on large clinical data sets show the effectiveness of the proposed architectures and demonstrate that the proposed algorithms can overcome the disadvantages mentioned above. Moreover, the proposed CNN models are fully automatic in terms of not requiring the extraction of diseased tissue, which is a great improvement of available automatic methods in the literature. To the best of the author's knowledge, this study is the first study to detect COVID-19 disease from given chest X-ray images, using CNN, whose hyperparameters are automatically determined by the Grid Search. Another important contribution of this study is that it is the first CNN-based COVID-19 chest X-ray image classification study that uses the largest possible clinical data set. A total of 1,524 COVID-19, 1,527 pneumonia, and 1524 normal X-ray images are collected. It is aimed to collect the largest number of COVID-19 X-ray images that exist in the literature until the writing of this research paper.

Chest CT for early detection and management of coronavirus disease (COVID-19): a report of 314 patients admitted to Emergency Department with suspected pneumonia.

PURPOSE: The purpose of our study was to assess the potential role of chest CT in the early detection of COVID-19 pneumonia and to explore its role in patient management in an adult Italian population admitted to the Emergency Department. METHODS: Three hundred and fourteen patients presented with clinically suspected COVID-19, from March 3 to 23, 2020, were evaluated with PaO2/FIO2 ratio from arterial blood gas, RT-PCR assay from nasopharyngeal swab sample and chest CT. Patients were classified as COVID-19 negative and COVID-19 positive according to RT-PCR results, considered as a reference. Images were independently evaluated by two radiologists blinded to the RT-PCR results and classified as "CT positive" or "CT negative" for COVID-19, according to CT findings. RESULTS: According to RT-PCR results, 152 patients were COVID-19 negative (48%) and 162 were COVID-19 positive (52%). We found substantial agreement between RT-PCR results and CT findings (p < 0.000001), as well as an almost perfect agreement between the two readers. Mixed GGO and consolidation pattern with peripheral and bilateral distribution, multifocal or diffuse abnormalities localized in both upper lung and lower lung, in association with interlobular septal thickening, bronchial wall thickening and air bronchogram, showed higher frequency in COVID-positive patients. We also found a significant correlation between CT findings and patient's oxygenation status expressed by PaO2/FIO2 ratio. CONCLUSION: Chest CT has a useful role in the early detection and in patient management of COVID-19 pneumonia in a pandemic. It helps in identifying suspected patients, cutting off the route of transmission and avoiding further spread of infection.

A promising approach for screening pulmonary hypertension based on frontal chest radiographs using deep learning: A retrospective study.

BACKGROUND: To date, the missed diagnosis rate of pulmonary hypertension (PH) was high, and there has been limited development of a rapid, simple, and effective way to screen the disease. The purpose of this study is to develop a deep learning approach to achieve rapid detection of possible abnormalities in chest radiographs suggesting PH for screening patients suspected of PH. METHODS: We retrospectively collected frontal chest radiographs and the pulmonary artery systolic pressure (PASP) value measured by Doppler transthoracic echocardiography from 762 patients (357 healthy controls and 405 with PH) from three institutes in China from January 2013 to May 2019. The wohle sample comprised 762 images (641 for training, 80 for internal test, and 41 for external test). We firstly performed a 8-fold cross-validation on the 641 images selected for training (561 for pre-training, 80 for validation), then decided to tune learning rate to 0.0008 according to the best score on validation data. Finally, we used all the pre-training and validation data (561+80 = 641) to train our models (Resnet50, Xception, and Inception V3), evaluated them on internal and external test dataset to classify the images as having manifestations of PH or healthy according to the area under the receiver operating characteristic curve (AUC/ROC). After that, the three deep learning models were further used for prediction of PASP using regression algorithm. Moreover, we invited an experienced chest radiologist to classify the images in the test dataset as having PH or not, and compared the prediction accuracy performed by deep learing models with that of manual classification. RESULTS: The AUC performed by the best model (Inception V3) achieved 0.970 in the internal test, and slightly declined in the external test (0.967) when using deep learning algorithms to classify PH from normal based on chest X-rays. The mean absolute error (MAE) of the best model for prediction of PASP value was smaller in the internal test (7.45) compared to 9.95 in the external test. Manual classification of PH based on chest X-rays showed much lower AUCs compared to that performed by deep learning models both in the internal and external test. CONCLUSIONS: The present study used deep learning algorithms to classify abnormalities suggesting PH in chest radiographs with high accuracy and good generalizability. Once tested prospectively in clinical settings, the technology could provide a non-invasive and easy-to-use method to screen patients suspected of having PH.

Advanced imaging tools for childhood tuberculosis: potential applications and research needs.

Tuberculosis is the leading cause of death globally that is due to a single pathogen, and up to a fifth of patients with tuberculosis in high-incidence countries are children younger than 16 years. Unfortunately, the diagnosis of childhood tuberculosis is challenging because the disease is often paucibacillary and it is difficult to obtain suitable specimens, causing poor sensitivity of currently available pathogen-based tests. Chest radiography is important for diagnostic evaluations because it detects abnormalities consistent with childhood tuberculosis, but several limitations exist in the interpretation of such results. Therefore, other imaging methods need to be systematically evaluated in children with tuberculosis, although current data suggest that when available, cross-sectional imaging, such as CT, should be considered in the diagnostic evaluation for tuberculosis in a symptomatic child. Additionally, much of the understanding of childhood tuberculosis stems from clinical specimens that might not accurately represent the lesional biology at infection sites. By providing non-invasive measures of lesional biology, advanced imaging tools could enhance the understanding of basic biology and improve on the poor sensitivity of current pathogen detection systems. Finally, there are key knowledge gaps regarding the use of imaging tools for childhood tuberculosis that we outlined in this Personal View, in conjunction with a proposed roadmap for future research.

OPTimal IMAging strategy in patients suspected of non-traumatic pulmonary disease at the emergency department: chest X-ray or ultra-low-dose chest CT (OPTIMACT) trial-statistical analysis plan.

BACKGROUND: A chest X-ray is a standard imaging procedure in the diagnostic work-up of patients suspected of having non-traumatic pulmonary disease. Compared to a chest X-ray, an ultra-low-dose (ULD) chest computed tomography (CT) scan provides substantially more detailed information on pulmonary conditions. To what extent this translates into an improvement in patient outcomes and health care efficiency is yet unknown. The OPTimal IMAging strategy in patients suspected of non-traumatic pulmonary disease at the emergency department: chest X-ray or ultra-low-dose chest CT (OPTIMACT) study is a multicenter, pragmatic, non-inferiority randomized controlled trial designed to evaluate replacement of chest X-ray by ULD chest CT in the diagnostic work-up of such patients, in terms of patient-related health outcomes and costs. During randomly assigned periods of 1 calendar month, either conventional chest X-ray or ULD chest CT scan was used as the imaging strategy. This paper presents in detail the statistical analysis plan of the OPTIMACT trial, developed prior to data analysis. METHODS/RESULTS: Functional health at 28 days is the primary clinical outcome. Functional health at 28 days is measured by the physical component summary scale of the Short Form (SF)-12 questionnaire version 1. Secondary outcomes are mental health (mental component summary scale of the SF-12), length of hospital stay, mortality within 28 days, quality-adjusted life year equivalent during the first 28 days (derived from the EuroQol five-dimension, five-level instrument), correct diagnoses at emergency department discharge as compared to the final post hoc diagnosis at day 28, number of patients in follow-up because of incidental findings on chest X-ray or ULD chest CT, and health care costs. CONCLUSIONS: After this pragmatic trial we will have precise estimates of the effectiveness of replacing chest X-ray with ULD chest CT in terms of patient-related health outcomes and costs. TRIAL REGISTRATION: Netherlands National Trial Register: NTR6163. Registered on 6 December 2016.